For an individual security calculating it's Sharpe and Sortino ratios is straightforward.

What I'm curious about is the following:

Let's say I have a portfolio of several securities, which is a distribution of my total capital: for example Asset A has 25%, Asset B has 50%, and Asset C has 25%. At every timestep t, let's assume that I can adjust these percentages to maximize my profits, and that the total distribution always has to add up to 100%.

So at each timestep t my portfolio has a return of r_t, which is the dot product of the distribution vector (a) for each asset at time t with the vector of the change in price for each asset since time t-1.

If I want to calculate the Sharpe and Sortino for the portfolio, would I:

  • Calculate the Sharpe and Sortino ratios for each individual security at time t and again take a dot product between my distribution vector a and the vector of each sharpe/sortino ratio for each security
  • Directly calculate the Sharpe and Sortino ratios of the portfolio using the returns of the portfolio (r_t) across all timesteps t.

Another good question would be: are both of these approaches fundamentally the same?

Thanks in advance for your help!


If you want to calculate the Sharpe and Sortino ratios for the portfolio, you should

  • directly calculate them using the returns of the portfolio

Even if the individual sharpe ratios for each of the $N$ assets being dot-multiplied by the portfolio weights is equivalent to the above approach, you would be calculating $N$ number of Sharpe/Sortino ratios when you could have just calculated the one that you want: the portfolio's ratio

| improve this answer | |
  • $\begingroup$ Could you clarify if ultimately the numerical result of both approaches is the same? I do understand that computationally your answer would be the most efficient. $\endgroup$ – Alex Pilafian Aug 23 at 18:54
  • 1
    $\begingroup$ you should try to program it to confirm whether they are equivalent approaches, but I would expect so, given that the portfolio performance ratio is computed from already aggregated (matrix) returns data, that is, the individual asset returns pre-weighted before the portfolio ratio is computed, whereas the one-by-one asset performance ratio approach, while not pre-weighted, is going at the same procedure except the weight vector is dot-applied, vector algebraically, to the sum of each individual asset's ratio. the redundancy for doing this will be why no one will even bother trying or respond $\endgroup$ – develarist Aug 23 at 19:02
  • $\begingroup$ This paper "Incremental Sharpe and other performance ratios" Benhamou and Guez (2018) explicitly states that the two methods produce the same result $\endgroup$ – Alex Pilafian Aug 24 at 13:26

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.